FLACCL: CUDA-enabled FLAC encoder by Gregory S. Chudov (prev. FlaCuda), Formerly "lossless codecs and CUDA" |
![]() ![]() |
FLACCL: CUDA-enabled FLAC encoder by Gregory S. Chudov (prev. FlaCuda), Formerly "lossless codecs and CUDA" |
Jan 22 2012, 10:40
Post
#301
|
|
|
Group: Members Posts: 32 Joined: 24-November 08 Member No.: 63054 |
QUOTE c:\Downloads\FLACCL_0.4\FLACCL>CUETools.FLACCL.cmd.exe --verify test.wav FLACCL#0.4, Copyright © 2010 Gregory S. Chudov. This is free software under the GNU GPLv3+ license; There is NO WARRANTY, to the extent permitted by law. <http://www.gnu.org/licenses/> for details. Filename : test.wav File Info : 44100kHz; 2 channel; 16 bit; 00:09:40.1600000 Error : unsupported residual coding Catalyst 12.1 Preview. This post has been edited by Miguk: Jan 22 2012, 10:40 |
|
|
|
Mar 18 2012, 16:17
Post
#302
|
|
|
Group: Members Posts: 2 Joined: 18-March 12 Member No.: 97886 |
Hey guys, I've been having MD5 signature mismatches with FlacCL on one of my computers with a nVidia GTX 570 card. A second computer, with a slower GT 430 card, works fine. Even with the signature mismatches, files (appear) to be fine, but I'm guessing something will be wrong, somewhere.
I wanted to do some troubleshooting on my own so that it doesn't sound like I'm just punting this out there. Hardware = Core i7 950, nVidia GTX 570, 12GB RAM Troubleshooting steps: FlacCL Level 11 - MD5 signature mismatch (4+ attempts, works ~10% of the time) (Speed = 235x) FlacCL Level 10 - MD5 signature mismatch (Speed = 254x) FlacCL Level 9 - No problems (try #1) / MD5 signature mismatch (try #2) (Speed = 281x) FlacCL Level 8 - MD5 signature mismatch (Speed = 325x) libFlake Level 11 - No problems (2 attempts) (Speed = 31x) flake Level 12 - No problems (Speed = 13x) FlacCL Level 11 - separate computer w/older nVidia card - no problems (Speed = 60x) File sizes don't even match for a full CD, but they are fairly close. (within a few kilobytes) Next step - Turned Verify on for FlacCL Level 11 (Speed = 136x) Got an "Exception: validation failed: iFrame=24603, ch=0" - it will be in a different spot each time. Sometimes it works. Next step - Tried Changing DoRice, GroupSize, and TaskSize. No consistent results. Worked a few times, failed most of the time. Next step - Change GPUOnly to False, and changed threads to 4, left Verify On (Speed = 308x) Haven't gotten this to fail. I've done 10 encodes of the same CD now. flac -t works every time. Next step - GPUOnly back to True, but leave threads at 4. Fails sometimes. GPUOnly=False seems to be required for it to work for me. So, how much should I worry about this? GPU Bad? Bug in CUETools with this card? Why would it work when I split the load with the GPU and the CPU? I've tried multiple nVidia drivers, including the latest. I have no problems leaving this the way it is, certainly if it works - it's faster than GPU only. If there is any other debugging information you want me to collect, I'd be happy to do that. agressiv |
|
|
|
Mar 18 2012, 20:17
Post
#303
|
|
![]() Group: Members Posts: 840 Joined: 7-October 01 Member No.: 235 |
I use FlacCl more or less from day 1 on, first with a GTX260 and now a GTX560Ti. I tried to provoke encoding errors with insane overclocking on the GPU and its RAM but couldn´t create a single one with reaching clockspeeds no game would run.
So my guess is you may have another stability problem with your CPU/Mainboard/GPU combination. You may try some stability test overclockers use. There only were 2 songs i found since Gregory created these GPU encoders that made problems but i don´t remember the exact errorcode back then. These were outright fixed by Gregory a while back. |
|
|
|
Mar 31 2012, 10:41
Post
#304
|
|
|
Group: Members Posts: 32 Joined: 24-November 08 Member No.: 63054 |
c:\CUETools>CUETools.FLACCL.cmd.exe --verify test.wav
FLACCL#0.4, Copyright © 2010 Gregory S. Chudov. This is free software under the GNU GPLv3+ license; There is NO WARRANTY, to the extent permitted by law. <http://www.gnu.org/licenses/> for details. Filename : test.wav File Info : 44100kHz; 2 channel; 16 bit; 00:09:40.1600000 Error : unsupported residual coding AMD Catalyst 12.3. |
|
|
|
Mar 31 2012, 10:52
Post
#305
|
|
![]() Group: Developer Posts: 648 Joined: 2-October 08 From: Ottawa Member No.: 59035 |
Does this happen on some specific input file, or does it happen on lots of files?
-------------------- CUETools 2.1.4
|
|
|
|
Apr 12 2012, 20:33
Post
#306
|
|
|
Group: Members Posts: 32 Joined: 24-November 08 Member No.: 63054 |
QUOTE c:\CUETools>CUETools.FLACCL.cmd.exe --verify "c:\_Music\Stefano Di Battista - Vo lare\new\01. Prima Di Partire.wav" FLACCL#0.4, Copyright © 2010 Gregory S. Chudov. This is free software under the GNU GPLv3+ license; There is NO WARRANTY, to the extent permitted by law. <http://www.gnu.org/licenses/> for details. Filename : c:\_Music\Stefano Di Battista - Volare\new\01. Prima Di Partire.wav File Info : 44100kHz; 2 channel; 16 bit; 00:05:32.5330000 Error : unsupported residual coding c:\CUETools>CUETools.FLACCL.cmd.exe --verify "c:\_Music\Stefano Di Battista - Vo lare\new\02. Blues For Michel.wav" FLACCL#0.4, Copyright © 2010 Gregory S. Chudov. This is free software under the GNU GPLv3+ license; There is NO WARRANTY, to the extent permitted by law. <http://www.gnu.org/licenses/> for details. Filename : c:\_Music\Stefano Di Battista - Volare\new\02. Blues For Michel.wav File Info : 44100kHz; 2 channel; 16 bit; 00:03:55.8400000 Error : unsupported residual coding c:\CUETools>CUETools.FLACCL.cmd.exe --verify "c:\_Music\Stefano Di Battista - Vo lare\new\03. Widows Walk.wav" FLACCL#0.4, Copyright © 2010 Gregory S. Chudov. This is free software under the GNU GPLv3+ license; There is NO WARRANTY, to the extent permitted by law. <http://www.gnu.org/licenses/> for details. Filename : c:\_Music\Stefano Di Battista - Volare\new\03. Widows Walk.wav File Info : 44100kHz; 2 channel; 16 bit; 00:09:22.0270000 Error : unsupported residual coding QUOTE c:\CUETools>flac -t "c:\_Music\Stefano Di Battista - Volare\new\02. Blues For Mi chel.flac" flac 1.2.1, Copyright © 2000,2001,2002,2003,2004,2005,2006,2007 Josh Coalson flac comes with ABSOLUTELY NO WARRANTY. This is free software, and you are welcome to redistribute it under certain conditions. Type `flac' for details. 02. Blues For Michel.flac: *** Got error code 0:FLAC__STREAM_DECODER_ERROR_STATU S_LOST_SYNC 02. Blues For Michel.flac: *** Got error code 0:FLAC__STREAM_DECODER_ERROR_STATU S_LOST_SYNC 02. Blues For Michel.flac: *** Got error code 0:FLAC__STREAM_DECODER_ERROR_STATU S_LOST_SYNC 02. Blues For Michel.flac: *** Got error code 0:FLAC__STREAM_DECODER_ERROR_STATU S_LOST_SYNC 02. Blues For Michel.flac: *** Got error code 0:FLAC__STREAM_DECODER_ERROR_STATU S_LOST_SYNC 02. Blues For Michel.flac: *** Got error code 0:FLAC__STREAM_DECODER_ERROR_STATU S_LOST_SYNC 02. Blues For Michel.flac: *** Got error code 0:FLAC__STREAM_DECODER_ERROR_STATU S_LOST_SYNC 02. Blues For Michel.flac: *** Got error code 0:FLAC__STREAM_DECODER_ERROR_STATU S_LOST_SYNC 02. Blues For Michel.flac: *** Got error code 0:FLAC__STREAM_DECODER_ERROR_STATU S_LOST_SYNC 02. Blues For Michel.flac: *** Got error code 0:FLAC__STREAM_DECODER_ERROR_STATU S_LOST_SYNC 02. Blues For Michel.flac: *** Got error code 0:FLAC__STREAM_DECODER_ERROR_STATU S_LOST_SYNC 02. Blues For Michel.flac: *** Got error code 0:FLAC__STREAM_DECODER_ERROR_STATU S_LOST_SYNC 02. Blues For Michel.flac: ERROR while decoding data state = FLAC__STREAM_DECODER_READ_FRAME AudioTester: QUOTE 3 files scanned in 0.12 seconds
--- 3 files failed C:\_Music\Stefano Di Battista - Volare\new\01. Prima Di Partire.flac (LOST_SYNC @ 0m 00s) C:\_Music\Stefano Di Battista - Volare\new\02. Blues For Michel.flac (LOST_SYNC @ 0m 00s) C:\_Music\Stefano Di Battista - Volare\new\03. Widows Walk.flac (LOST_SYNC @ 0m 00s) --- 0 files passed This post has been edited by Miguk: Apr 12 2012, 20:37 |
|
|
|
Apr 27 2012, 19:46
Post
#307
|
|
|
Group: Members Posts: 3 Joined: 27-April 12 Member No.: 99263 |
I'm also getting the same error. It happens on pretty much every file I throw at it (although some worked fine, I think), ever since I switched from an NVidia GPU to an ATI one (although I used FlaCuda with the NVidia one, so that's probably not of much help). Oh, and it works just fine if I tell it to encode on the CPU only, instead of the GPU, though, of course, much slower.
|
|
|
|
Apr 27 2012, 19:52
Post
#308
|
|
![]() Group: Developer Posts: 648 Joined: 2-October 08 From: Ottawa Member No.: 59035 |
Please tell which exactly GPU do you use. And if possible, driver version as well.
-------------------- CUETools 2.1.4
|
|
|
|
Apr 27 2012, 22:46
Post
#309
|
|
|
Group: Members Posts: 3 Joined: 27-April 12 Member No.: 99263 |
Right, I meant to include that, but must've forgotten.
It's a PowerColor Radeon HD 6850. The current driver is the latest one, 8.961.0.0, the one that came with the latest release just a few days back. Catalyst 12.4, to be exact. But it wasn't working with at least 3 previous releases of the drivers as well, if it helps you. Don't know if it worked with earlier releases, as I never tried the earlier ones. |
|
|
|
Apr 27 2012, 22:46
Post
#310
|
|
![]() Group: Developer Posts: 648 Joined: 2-October 08 From: Ottawa Member No.: 59035 |
Thanks
-------------------- CUETools 2.1.4
|
|
|
|
Apr 28 2012, 18:28
Post
#311
|
|
|
Group: Members Posts: 32 Joined: 24-November 08 Member No.: 63054 |
I am using Radeon HD 5870.
As I mentioned earlier, on my Nvidia ION with the latest Nvidia driver FLACCL doesn't work, too. But it works with 267.54 driver. |
|
|
|
May 24 2012, 06:05
Post
#312
|
|
|
Group: Members Posts: 6 Joined: 20-November 06 Member No.: 37813 |
Having the same "Error : unsupported residual coding" in CueTools. Running an ATI 6950 with Catalyst 12.4 and the latest OpenCL SDK
|
|
|
|
May 29 2012, 09:00
Post
#313
|
|
![]() Server Admin Group: Admin Posts: 4808 Joined: 24-September 01 Member No.: 13 |
|
|
|
|
May 29 2012, 14:13
Post
#314
|
|
![]() Group: Developer Posts: 648 Joined: 2-October 08 From: Ottawa Member No.: 59035 |
Very interesting. Thanks a lot!
I'm not entirely surprised that SLI configurations and dual chip cards are often slower than single chip, not all algorithms benefit from this, especially without special fine-tuning. But it's surprising how Kepler seems slower than Fermi. In theory it should have outperformed AMD. I will have to investigate what's going on here. -------------------- CUETools 2.1.4
|
|
|
|
May 29 2012, 14:25
Post
#315
|
|
|
Group: Developer (Donating) Posts: 2041 Joined: 19-October 01 From: Finland Member No.: 322 |
On my PCIe 3.0 configuration SLI with two GTX 680 cards was only 2% slower than encoding with one card inside. The difference between SLI on and off when two cards were in the machine was about 0.1%.
|
|
|
|
May 29 2012, 14:35
Post
#316
|
|
|
Group: Members Posts: 32 Joined: 24-November 08 Member No.: 63054 |
|
|
|
|
May 29 2012, 14:52
Post
#317
|
|
|
Group: Members Posts: 1315 Joined: 3-January 05 From: Argentina, Bs As Member No.: 18803 |
Interesting. I wonder what performance is per Watt for CPU vs GPU (especially for mobile AMD APU). This post has been edited by IgorC: May 29 2012, 14:53 |
|
|
|
Jun 15 2012, 13:32
Post
#318
|
|
![]() Group: Members Posts: 43 Joined: 23-May 08 From: Rzeczpospolita Member No.: 53744 |
Intel® Core™ i7-3770, 8Gb DDRIII-1333, Win7 HP x64
CUETools.FLACCL.cmd.exe -o nul --cpu-threads 8 --opencl-type CPU --opencl-platform "Intel® OpenCL" test.wav -0 = 1552.67 -1 = 1402.22 -2 = 1308.66 -3 = 808.33 -4 = 645.38 -5 = 460.92 -6 = 654.20 -7 = 482.51 -8 = 278.95 -9 = 322.66 -10 = 186.75 -11 = 132.91 This post has been edited by no404error: Jun 15 2012, 13:32 |
|
|
|
Jun 19 2012, 14:43
Post
#319
|
|
|
Group: Members Posts: 32 Joined: 24-November 08 Member No.: 63054 |
I updated today my ION (notebook) video drivers to the latest NVIDIA version (301.42 x64).
FLACCL test failed again but due to other reasons: - test 1: only 6 of 9 tracks encoded - test 2: only 8 of 9 tracks encoded - audiotester showed 0 errors in encoded tracks in both cases - cuetools showed every time following error: ![]() flaclib encoded all tracks without any problem. This post has been edited by Miguk: Jun 19 2012, 14:46 |
|
|
|
Jun 20 2012, 15:33
Post
#320
|
|
|
Group: Members Posts: 32 Joined: 24-November 08 Member No.: 63054 |
flaclib = libflac :-)
|
|
|
|
Jun 23 2012, 19:02
Post
#321
|
|
|
Group: Members Posts: 1315 Joined: 3-January 05 From: Argentina, Bs As Member No.: 18803 |
Intel® Core™ i7-3770, 8Gb DDRIII-1333, Win7 HP x64 CUETools.FLACCL.cmd.exe -o nul --cpu-threads 8 --opencl-type CPU --opencl-platform "Intel® OpenCL" test.wav -0 = 1552.67 -1 = 1402.22 -2 = 1308.66 -3 = 808.33 -4 = 645.38 -5 = 460.92 -6 = 654.20 -7 = 482.51 -8 = 278.95 -9 = 322.66 -10 = 186.75 -11 = 132.91 The speed is so high for real use Can You run the same configuration but with --cpu-threads 1 or gpu only? It will give a rough idea what to expect from mobile CPUs. Thank You. This post has been edited by IgorC: Jun 23 2012, 19:03 |
|
|
|
Jun 23 2012, 19:16
Post
#322
|
|
![]() Group: Developer Posts: 648 Joined: 2-October 08 From: Ottawa Member No.: 59035 |
Can You run the same configuration but with --cpu-threads 1 or gpu only? It will give a rough idea what to expect from mobile CPUs. Actually, it won't. --cpu-threads option's name is probably a bit confusing in case of --opencl-type CPU. There are two types of threads - actual opencl threads, which run on GPU(s) in case of --opencl-type GPU, and run on CPU(s) in cases of --opencl-type CPU. Number of those threads is not controlled by --cpu-threads option, which controls the number of threads of second type - control threads, that schedule opencl work. So i should probably rename --cpu-threads to --control-threads and maybe introduce a new option --cpu-cores, to control number of CPU cores used by opencl tasks in case of --opencl-type CPU. -------------------- CUETools 2.1.4
|
|
|
|
Jun 23 2012, 20:13
Post
#323
|
|
![]() Group: Members Posts: 43 Joined: 23-May 08 From: Rzeczpospolita Member No.: 53744 |
Intel® Core™ i7-3770, 8Gb DDRIII-1333, Win7 HP x64
CUETools.FLACCL.cmd.exe -o nul --opencl-type GPU test.wav -0 = 375.28 -1 = 764.13 -2 = 1038.91 -3 = 314.01 -4 = 345.24 -5 = 303.78 -6 = 346.60 -7 = 304.84 -8 = 177.76 -9 = 120.67 -10 = 85.10 -11 = 66.12 |
|
|
|
Jun 27 2012, 18:59
Post
#324
|
|
|
Group: Members Posts: 82 Joined: 28-September 07 From: Petrozavodsk Member No.: 47418 |
Got gass.cuda.cudaexception on i7-2600k/GTX570 with 304.48 drivers. GTX280 worked fine.
|
|
|
|
Jul 31 2012, 01:19
Post
#325
|
|
|
Group: Members Posts: 3 Joined: 27-April 12 Member No.: 99263 |
Any update on the "Error : unsupported residual coding" error? Perhaps there are some test builds you'd like us (those who experience the error) to run?
|
|
|
|
![]() ![]() |
|
Lo-Fi Version | Time is now: 25th May 2013 - 01:30 |